Dr Chis Meah assesses business’s attitudes to artificial intelligence and provides a few pointers as to how best to approach it.
There’s this thing called AI - maybe you’ve heard of it?
Since late 2022, with the launch of ChatGPT, the growth in interest has become AI mania. “Artificial” and “intelligence” are the new Posh and Becks, it seems – I’ll let you choose which is which.
There’s just one slight problem: companies are making strategic decisions about AI when most people still don’t know what it actually is. It’s like you’re at a party, talking to someone for a while, and you realise you don’t know their name. It’s too late to ask. You then find out that no one knows their name. Everyone is just nodding along, calling them “mate”, with no clue what their name actually is.
That’s current AI in a nutshell. It’s fine for parties, but not business strategies. Boards are demanding an AI plan. Customers expected “AI‑powered” features yesterday. Your team swings between “Have you ChatGPT’d it?” and AI‑fuelled nightmares.
Committing to an AI strategy without understanding it is like marrying the party guest whose name you never asked – and AI providers don’t have a pre-nup in their Terms and Conditions.
The advice I give to leaders and organisations is to get literate in what AI is, ask what it can do for you, and think where it might be going.
A working definition
For practical purposes, treat the current state of AI as: software that predicts and generates language (and other patterns) from examples. That turns out to be incredibly powerful!
Predicting the next word in a sentence unlocks lots of use cases and opportunities. But, there’s no Wizard of Oz behind the curtain; that is all that’s happening. We could dive deeper into understanding, but let’s examine what it enables us to do.
What you can do now (that actually helps)
Today, most online meetings have an AI assistant that can automatically record, transcribe, summarise, and even extract and assign actions from the conversation. Magical!
You can also use those same tools offline. In face-to-face meetings, you can open your Zoom, Notion, or most other meeting apps on your phone and record the meeting in the same way. You can meet in person, but have all the automated diligence of your online workflow.
But we can make AI an active participant. You have a board meeting. You decide on an exciting new direction. Everyone is ecstatic. They're high-fiving. Could AI be the “sceptical 10th person”? If “9 out of 10 people agree”, what would the 10th person say? AI could digest the conversation, extract information from across your company, and give three reasons why the direction might fail. That raises decision quality, robustness and speed now - no need to wait for AI improvements or sci-fi futures.
What you can’t do now (without a lot of pain)
There’s much talk about the end of software development and AI agents writing all the code. “Vibe coding” is trending. But like the Pokémon cards, pogs, and pet rocks that came before, the craze has narrow and limited uses.
If you want some intuition for the current limitations, consider this: the technology we use for producing words (in things like ChatGPT) can also be used to predict the weather.
If you want to predict today’s weather, it’s likely to be highly accurate! Tomorrow’s too. What about the weather next month? Next year? Thirty years from now?
You intuitively know that the further out you have to predict the weather, the less likely you are to get an accurate forecast. We can forecast later today, and then use that to forecast tomorrow, but as you chain these predictions together, you will drift further from reality. In 30 years, there are so many factors that you will be assuming incorrectly that the prediction will be gibberish – a shot in the dark.
That’s what happens with AI coding (and most complex AI tasks) right now. For a simple task, it is excellent! You can knock up a complete and rapid prototype easily – that has incredible value if harnessed properly. However, solving a complex problem with a maintainable system requires a coherent thought process – something that our current AI doesn’t have. Instead, for now, AI in these complex scenarios can augment human capacity, but cannot replace it entirely.
How to evaluate (the messy 80%)
My rule of thumb: AI gets you 80% there – but the path to 100% finished product isn’t a straight line. You’ll revisit messy ground because AI will have done things in a way you cannot always understand.
It’s like if you were aiming for a 100-floor skyscraper, AI would build to floor 80 but use jelly instead of cement.
Where is 80% good enough?
So, is it game over for AI at this point? No. It’s incredible. In generative AI and large language models, we’ve created a way to communicate fluently with computers in plain language. That is amazing, and makes this AI revolution the most accessible in history. If you can use language, you can interact with these systems.
But you need to know how and where to use it. Ask yourself, “Where is 80% good enough?” There’s definitely a place, whether you’d like to admit it or not.
Take food. Michelin star restaurants are the pinnacle – 100% food. Then there’s McDonald’s. It feeds many more people, with 1,270 locations in the UK versus 188 Michelin-starred venues, usually with queues around the corner of any drive-through. I’d call it 80% food.
Is there some ‘McDonaldsification’ you can bring to areas of your business, so that you can free your team up to work on what needs to be Michelin Star quality? Look for places where your people don’t enjoy things, where you need to achieve at a large scale, or where you don’t currently add value but could with the help of AI.
AI moves at a rapid pace. Do you have to keep up with every development? It’s impossible – the pace of change is too high to keep track of all the thousands of shiny new AI-empowered apps that come out each week. The solution is to ask: do you have to be the hare, or can you be the tortoise?
Where you can, pick one general tool and use it. These include offerings from OpenAI with ChatGPT, Anthropic with Claude, Google with Gemini, and X with Grok. These models will do most jobs well enough, and you won’t need to keep switching. If you can afford to wait around six months, these general models will likely absorb today’s specific tools' niche features. If you can, be a tortoise.
If you can’t wait — because the capability is core to your competitive advantage — be a hare. Try the specialist tools, benchmark them all, pick a winner, and stay updated so you can keep at the fast-moving frontier. But you can only sprint in two or three directions before you spin yourself silly, so limit it to what drives your specific advantage.
The key is whether you’re the tortoise or the hare; you’re still in the race. You want to build a culture of using and experimenting with AI now, and move forward. As the tools improve, you’ll be running and ready to adopt, rather than standing on the starting blocks.
AI can feel scary to many people. Let’s be clear – without team buy-in, any AI attempts are destined to fail! There is no other way to make this work long-term. Many companies have been short-sighted in laying off staff they don’t think they will need, and they may well pay the financial and reputational price in the future.
How do you get team buy-in? Show them the value: both theirs and AIs.
You have to get past identifying with job roles – if people do that, then any automation of that role is a threat. Instead, you value people because they know the culture, they have the relationships, and they have your trust.
If they can automate some of their tasks, then great! You can support them in doing more for the company in other ways. They have to understand the good on the other side. Show you value them, have clear communication, and reward the right behaviours.
Then, have people see the benefits of AI. Remember Marie Kondo? Lay out all your tasks, see what sparks joy or value, and keep it! Automate the rest with AI, or delete the non-valuable tasks. AI will help your people focus and be more effective in what they enjoy and add the most value to.
AI can help you in many ways, but focus on two: What can AI help you do quicker, easier, or cheaper? What can AI help you do that you could never do before, or do in a new way?
Where might it go (and where faith kicks in)
Will AI handle complex, multi-step work end-to-end? Possibly. The intoxicating promise is that with more data and compute power, we get more intelligence. But, at the moment, there are three dots on a whiteboard, and we’re leaping to “everyone can see that dog I’ve drawn, right?”
The bridge from today’s systems to general intelligence includes a leap of faith.
Near‑term progress is more likely to be very practical: tools that call tools (APIs or MCPs), richer memory with your internal data, and multiple specialised ‘agents’ collaborating under human supervision. That’s plenty to transform operations without waiting for the marketing hype and AGI promises to become reality.
What might change (and what it rewards)
Revenue per employee becomes the quality metric.
Revenue alone won’t be the focus. Consider which business you would prefer to own or invest in: £100 million revenue with 10,000 employees is £10,000 per employee. £25 million with 100 employees is £250,000 per employee—25 times the leverage. AI pushes firms to ask: how much value can each person create with the right tools, data, and environment? Revenue per employee will be the standard metric.
Culture shifts from activity to outcomes.
To embrace the opportunity, you need a strong culture to be an asset in your business. Let’s describe an asset as something that, if you went travelling the world for six months, would still be there delivering value when you return. If your culture relies on you, it’s not an asset. It depends on your continued presence, time and effort. You need to build a sustainable culture which prioritises learning, focuses on experimentation, and measures progress in outcomes.
What stays the same (the non‑negotiables)
What stays the same? I run AI Retreats for CEOs and leaders. We explore AI from top to bottom: what it is, what it means, where it’s going. All of that is important, but we always tie everything back to the company’s vision.
If you had a magic wand and anything was possible, what would your company do? How would it solve problems for customers? What would the team and operations look like? Where would people spend their time? Having a clear and compelling vision makes AI decisions easy – does it help us towards the vision?
Yes? Experiment away.
No? It’s a distraction.
Audience and distribution
In a world where anyone can make anything, the key is distribution. An audience that trusts you is the advantage. ‘Influencers’ and content creators get this (and it’s often all they need).
AI will increasingly squash the distance between idea and execution, and if creation is easy then the audience is king! You cater to them, and they reward you with trust. A group of people who trust you will choose to buy from you. Creation is being commoditised; trust and access are power.
Final word
AI is powerful. It will get better.
However, there is a lot of hype to dodge, and a lot of uncertainty lies ahead.
Keep an open mind, get value from the tools we have now, and build a culture that can thrive in uncertainty.
Dr Chris Meah
